April 23, 2015 Mapping croplands using Landsat data with generalized classifier over large areas Aparna Phalke and Prof. Mutlu Ozdogan Nelson Institute for Environmental Studies University of Wisconsin - Madison
Updates on following Progress in processing random scenes Results on turkey and Syria Random scene processing
Methods for Generalization Create buffer of 915 km around the original footprint Select random three random footprints for each
Randomly Selected footprints
Landsat Footprints collected Random1Random2Random3 Selected Footprintsscenespathrowscenespathrowscenespathrowscenes Turkey(172/34) Syria(174/35) Tunisia(192/35) Algeria(193/35) Egypt(177/39) Morocco(201/37) Romania(182/29) Bulgaria(182/30) Ukraine(181/26) Poland(189/24) Germany(193/24) UK(202/24) France(199/26) Hungary(187/27) Italy(192/29) Portugal(203/33) Spain(200/33) Total All total2460 RandomTotal1772
Methodology 1. Atmospheric Correction2. Cloud Masking3. EVI creation4. EVI Stacking5. EVI Temporal Statistics6. Performing LDA Model7. Developing Crop/non-crop map8. Accuracy assessment
Methodology Accuracy of Random scenes Each scene trained by 100 random points These random points are used to get overall accuracy.
Results Random : Turkey 1, Turkey 2, Turkey 3
Accuracy 68.00% Accuracy 71.00% Accuracy 63.00%
Accuracy 89.00% Accuracy 86.00% Accuracy 88.00%
Accuracy 89.00% Accuracy 87.00% Accuracy 66.00%
Results Random : Syria 1, Syria 2, Syria 3
Accuracy 83.00% Accuracy 87.00% Accuracy 81.00%
Accuracy 98.00% Accuracy 64.00% Accuracy 93.00%
Accuracy 73.00% Accuracy 57.00% Accuracy 76.00%
Thank you